Explore the vast range of titles available.

Background The number of liver resections is constantly rising over the last decades. Despite the reduction of overall mortality and morbidity in liver surgery, biliary leakage is still a relevant postoperative complication that can lead to a fatal postoperative course. Aim of this analysis is the identification of specific risk factors for postoperative biliary complications after liver resections and the development of a predictive biliary leakage risk score. Methods A single-center, retrospective analysis of 844 liver resections performed in the Department of Visceral, Thoracic and Vascular Surgery, Technische Universität Dresden, between 1/2013 and 12/2019 is conducted to identify risk factors for postoperative biliary leakage and a risk score for biliary leakage after hepatectomy is established based on multivariate regression. The score has been validated by an independent validation cohort consisting of 142 patients. Results Overall morbidity is 43.1% with 36% surgical complications and an overall mortality of 4.3%. Biliary leakage occurred in 15.8% of patients. A predictive score for postoperative biliary leakage based on age, major resection, pretreatment with FOLFOX/cetuximab and operating time is created. Patients are stratified to low (< 15%) and high (> 15%) risk with a sensitivity of 67.4% and a specificity of 70.7% in development cohort and a specificity of 68.2% and sensitivity of 75.8% in validation cohort. Conclusions The presented score is robust and has been validated in an independent patient cohort. Depending on the calculated risk, prevention or early treatment can be initiated to avoid bile leakage and to improve postoperative course.

Type 1 diabetes begins with autoimmunity against pancreatic islet antigens, including insulin. The aim of the Primary Oral Insulin Trial (POInT) was to evaluate the efficacy and safety of daily high-dose oral insulin to prevent the development of islet autoantibodies and diabetes. In this randomised, controlled, primary prevention trial, genetic screening in seven obstetric and paediatric clinics in Germany, Poland, Sweden, Belgium, and the UK identified newborns with a greater than 10% risk of developing islet autoimmunity. Eligible infants aged 4–7 months were randomly assigned in a 1:1 ratio to receive insulin manufactured from human zinc–insulin crystals administered orally at a once-daily dose of 7·5 mg for 2 months, increasing to 22·5 mg for 2 months and 67·5 mg until age 3 years, or placebo. Participants were randomly assigned via a web-based application and were stratified by site. The primary outcome was the development of two or more islet autoantibodies or diabetes assessed throughout follow-up until a maximum age of 6·5 years. A secondary outcome was the development of dysglycaemia or diabetes. Islet autoantibodies were measured in samples collected at baseline and during study visits conducted at outpatient clinics at 2, 4, and 8 months after randomisation, at age 18 months, and every 6 months thereafter. All participants and their family members, investigators of the study, and laboratory personnel remained masked to treatment allocation during the whole study. All randomly assigned participants who correctly fulfilled eligibility criteria and had not reached the primary outcome at the baseline visit (modified intention-to-treat) were included in the primary analysis. All participants who received at least one dose of study drug were included in the safety analysis. POInT is registered with ClinicalTrials.gov (NCT03364868) and is complete. Of 241 977 screened newborns, 2750 (1·14%) had an elevated genetic risk of developing islet autoimmunity and 1050 (38·2%) of the eligible infants (531 males [51%], 519 females [49%]), were assigned to oral insulin or placebo between Feb 7, 2018, and March 24, 2021. Two participants in the oral insulin group and none in the placebo group were excluded from the modified intention-to-treat analysis. The primary outcome developed in 52 (10%) participants in the insulin group and 46 (9%) in the placebo group (hazard ratio 1·12 [95% CI 0·76–1·67], p=0·57). An interaction between treatment and the INS rs1004446 genotype was observed, with an increase in the primary outcome in participants in the insulin group carrying non-susceptible INS genotypes compared with the placebo group (2·10 [1·08–4·09]) and protection against diabetes or dysglycaemia in participants in the insulin group carrying susceptible INS genotypes compared with the placebo group (0·38 [0·17–0·86]). Blood glucose values less than 50 mg/dL were observed in two (0·03%) of 7210 measurements in the insulin group and six (0·08%) of 7070 measurements in the placebo group. Of 10 252 reported adverse events, 5076 (49·5%) occurred in 507 (96·0%) of 528 participants in the oral insulin group and 5176 (50·5%) occurred in 500 (95·8%) of 522 participants in the placebo group. One death occurred in the oral insulin group and was unrelated to the study drug following independent review. There was no evidence that high-dose, daily oral insulin prevents the development of islet autoantibodies. Further studies are needed to assess the benefit of primary oral insulin therapy for preventing diabetes in INS genotype-selected infants. Leona M and Harry B Helmsley Charitable Trust.

The ability to automatically convert natural language text into a knowledge base may open the door to great new opportunities, including question-answering on the Web, detection of trends and sentiments in social media, and perhaps even intelligent agents which understand our language. Today, however, there does not exist a system that can reliably convert text into a knowledge base, and the task turns out to be far more difficult than it appears. A key challenge is relation extraction – detecting semantic relationships between entities mentioned in text. Most successful approaches use supervised machine learning, but creating the required labeled training examples has proven too expensive for constructing Web-scale knowledge bases. This dissertation shows that we can greatly reduce the amount of human effort necessary to create relation extractors by leveraging a richer set of user interactions, some of which use more accurate models of weak supervision from a database. Specifically, this dissertation presents (1) a weakly supervised technique based on multi-instance learning which allows relations to overlap, (2) a weakly supervised technique that allows learning from only a few instances per relation by dynamically inducing relation-specific lexicons, (3) an approach for developing extraction rules interactively, and (4) a technique which synergistically pairs weakly supervised relation extraction with extraction validation by an online community. Our proposed techniques make it possible to create a high-quality relation extractor in under one hour, moving us closer towards automatically constructing Web-scale knowledge-bases.

We introduce EmbeddingGemma, a new lightweight, open text embedding model based on the Gemma 3 language model family. Our innovative training recipe strategically captures knowledge from larger models via encoder-decoder initialization and geometric embedding distillation. We improve model robustness and expressiveness with a spread-out regularizer, and ensure generalizability by merging checkpoints from varied, optimized mixtures. Evaluated on the Massive Text Embedding Benchmark (MTEB) across multilingual, English, and code domains, EmbeddingGemma (300M) achieves state-of-the-art results. Notably, it outperforms prior top models, both proprietary and open, with fewer than 500M parameters, and provides performance comparable to models double its size, offering an exceptional performance-to-cost ratio. Remarkably, this lead persists when quantizing model weights or truncating embedding outputs. This makes EmbeddingGemma particularly well-suited for low-latency and high-throughput use cases such as on-device applications. We provide ablation studies exploring our key design choices. We release EmbeddingGemma to the community to promote further research.

In this report, we introduce Gemini Embedding, a state-of-the-art embedding model leveraging the power of Gemini, Google's most capable large language model. Capitalizing on Gemini's inherent multilingual and code understanding capabilities, Gemini Embedding produces highly generalizable embeddings for text spanning numerous languages and textual modalities. The representations generated by Gemini Embedding can be precomputed and applied to a variety of downstream tasks including classification, similarity, clustering, ranking, and retrieval. Evaluated on the Massive Multilingual Text Embedding Benchmark (MMTEB), which includes over one hundred tasks across 250+ languages, Gemini Embedding substantially outperforms prior state-of-the-art models, demonstrating considerable improvements in embedding quality. Achieving state-of-the-art performance across MMTEB's multilingual, English, and code benchmarks, our unified model demonstrates strong capabilities across a broad selection of tasks and surpasses specialized domain-specific models.

Information Extraction (IE) aims to automatically generate a large knowledge base from natural language text, but progress remains slow. Supervised learning requires copious human annotation, while unsupervised and weakly supervised approaches do not deliver competitive accuracy. As a result, most fielded applications of IE, as well as the leading TAC-KBP systems, rely on significant amounts of manual engineering. Even \"Extreme\" methods, such as those reported in Freedman et al. 2011, require about 10 hours of expert labor per relation. This paper shows how to reduce that effort by an order of magnitude. We present a novel system, InstaRead, that streamlines authoring with an ensemble of methods: 1) encoding extraction rules in an expressive and compositional representation, 2) guiding the user to promising rules based on corpus statistics and mined resources, and 3) introducing a new interactive development cycle that provides immediate feedback --- even on large datasets. Experiments show that experts can create quality extractors in under an hour and even NLP novices can author good extractors. These extractors equal or outperform ones obtained by comparably supervised and state-of-the-art distantly supervised approaches.

Metabolomics using nontargeted tandem mass spectrometry can detect thousands of molecules in a biological sample. However, structural molecule annotation is limited to structures present in libraries or databases, restricting analysis and interpretation of experimental data. Here we describe CANOPUS (class assignment and ontology prediction using mass spectrometry), a computational tool for systematic compound class annotation. CANOPUS uses a deep neural network to predict 2,497 compound classes from fragmentation spectra, including all biologically relevant classes. CANOPUS explicitly targets compounds for which neither spectral nor structural reference data are available and predicts classes lacking tandem mass spectrometry training data. In evaluation using reference data, CANOPUS reached very high prediction performance (average accuracy of 99.7% in cross-validation) and outperformed four baseline methods. We demonstrate the broad utility of CANOPUS by investigating the effect of microbial colonization in the mouse digestive system, through analysis of the chemodiversity of different Euphorbia plants and regarding the discovery of a marine natural product, revealing biological insights at the compound class level. Unknown metabolites are classified from mass spectrometry data.

β-arrestins mediate regulatory processes for over 800 different G protein-coupled receptors (GPCRs) by adopting specific conformations that result from the geometry of the GPCR–β-arrestin complex. However, whether β-arrestin1 and 2 respond differently for binding to the same GPCR is still unknown. Employing GRK knockout cells and β-arrestins lacking the finger-loop-region, we show that the two isoforms prefer to associate with the active parathyroid hormone 1 receptor (PTH1R) in different complex configurations (“hanging” and “core”). Furthermore, the utilisation of advanced NanoLuc/FlAsH-based biosensors reveals distinct conformational signatures of β-arrestin1 and 2 when bound to active PTH1R (P-R*). Moreover, we assess β-arrestin conformational changes that are induced specifically by proximal and distal C-terminal phosphorylation and in the absence of GPCR kinases (GRKs) (R*). Here, we show differences between conformational changes that are induced by P-R* or R* receptor states and further disclose the impact of site-specific GPCR phosphorylation on arrestin-coupling and function. Here the authors present improved intramolecular sensors for β-arrestin2 and 1, which enable assessment of conformational changes of both isoforms in living cells. These reveal that the same GPCR induces differential conformational rearrangements that determine the functional diversity between the two β-arrestins.

The present study aimed to investigate the possible mechanisms linking a single-item measure of global self-rated health (SRH) with morbidity by comparing the association strengths between SRH with markers of autonomic nervous system (ANS) function, inflammation, blood glucose and blood lipids. Cross-sectional comprehensive health-check data of 3947 working adults (age 42±11) was used to calculate logistic regressions, partial correlations and compare correlation strength using Olkins Z. Adjusted logistic regression models showed a negative association between SRH (higher values indicating worse health) and measures of heart rate variability (HRV). Glycemic markers were positively associated with poor SRH. No adjusted association was found with inflammatory markers, BP or lipids. In both unadjusted and adjusted linear models Pearson's correlation strength was significantly higher between SRH with HRV measures compared to SRH with other biomarkers. This is the first study investigating the association of ANS function and SRH. We showed that a global measure of SRH is associated with HRV, and that all measures of ANS function were significantly more strongly associated with SRH than any other biomarker. The current study supports the hypothesis that the extent of brain-body communication, as indexed by HRV, is associated with self-rated health.